Sepsis. National Clinical Guideline Centre. Sepsis: the recognition, diagnosis and management of sepsis. NICE guideline <number> January 2016

Transcription

1 National Clinical Guideline Centre Consultation Sepsis Sepsis: the recognition, diagnosis and management of sepsis NICE guideline <number> Appendices I-O January 2016 Draft for consultation Commissioned by the National Institute for Health and Care Excellence

2

3 Contents Disclaimer Healthcare professionals are expected to take NICE clinical guidelines fully into account when exercising their clinical judgement. However, the guidance does not override the responsibility of healthcare professionals to make decisions appropriate to the circumstances of each patient, in consultation with the patient and, where appropriate, their guardian or carer. Copyright Funding National Institute for Health and Care Excellence

4 Contents Contents Appendices... 5 Appendix I: Economic evidence tables... 5 Appendix J: GRADE tables Appendix K: Forest plots Appendix L: Excluded clinical studies Appendix M:Excluded health economic studies Appendix N: Research recommendations Appendix O: NICE technical team References: Appendix I-O

5 Economic evidence tables 5 1 Appendices 2 Appendix I: Economic evidence tables I.1 3 Scoring systems 4 None. I.2 5 Signs and symptoms 6 None. I.3 7 Blood tests 8 None. I.4 9 Lactate 10 None. I.5 11 Serum creatinine 12 None. I.6 13 Disseminated intravascular coagulation 14 None.

6 Economic evidence tables 6 I.7 15 Antimicrobial treatment 16 None. I.8 17 IV fluid administration 18 None. I.9 19 Escalation of care 20 None. I Inotropic agents and vasopressors 22 None. I Supplemental oxygen 24 None. I Use of bicarbonate 26 None. I Early goal-directed therapy (EGDT) 28 Study Mouncey Study details Population & interventions Costs Health outcomes Cost effectiveness Economic analysis: CUA (health outcome: QALYs) Population: Patients with early signs of Total costs (mean per patient): QALYs (mean per patient): Intervention 1: ICER (Intervention 2 versus Intervention 1): Intervention 2 dominated (more expensive

7 Economic evidence tables 7 Study design: Within trial analysis (RCT) Approach to analysis: Analysis of individual level data for mortality and EQ- 5D. Unit costs were applied to resource use. Perspective: UK NHS Time horizon/follow-up 90 days QoL follow up Treatment effect duration: Resuscitation protocol was followed for 6 hours Discounting: Costs: NR; Outcomes: NR septic shock Patient characteristics: N = 1251 Mean age: invtn 1 = 64.3 (15.5), intvn 2 = 66.4 (14.6) Male: invtn 1 = 58.6%, intvn 2 = 57% Intervention 1: Usual care The usual care group continued to receive monitoring, investigation and treatment as determined by the clinician. Intervention 2: Early Goal Directed Therapy (EGDT). Following a resuscitation protocol involving central venous catheter insertion with central venous oxygen saturation monitoring capability and intensive therapy of other interventions Intervention 1: 11,424 Intervention 2: 12,414 Incremental (2 1): 989 (95% CI: -726 to 2,705; p=nr) Currency & cost year: 2012 UK pounds Cost components incorporated: - Equipment and consumables 2 monitors capable of oxygen saturation monitoring assumed to be needed per hospital. Costs of consumables including the catheter capable of monitoring, pressure transducers. - Blood products and dobutamine - Staff time to deliver the protocol; time for vascular catheter insertion and time for monitoring patients (assumed 10 minutes of nurse time per hr of the resus protocol). Staff time for training, assumed to be 20 minutes per ED staff member every 5 years (5 years assumed to be the life Intervention 2: Incremental (2 1): (95% CI: to 0.005); p=0.85) and less benefit) Probability Intervention 2 cost-effective ( 20K/30K threshold): 12%/12% (read from graph) Analysis of uncertainty: Some form of PSA undertaken (a) to generate cost effectiveness plane and cost effectiveness acceptability curve. 500 estimates obtained. Sensitivity analyses undertaken include: - Manufacturer list price used for monitoring machines instead of discounted price used in base case - Staff monitoring time varied from 10 minutes per hour in the base case to 5 and 15 minutes. - Location of protocol implementation; if protocol is implemented in the ED, staffneed to be trained but in critical care they do not. Sensitivity analysis assumed that the protocol was implemented either exclusively in the ED or critical care. - Re-admission data in the base case was gathered both from the health services questionnaire sent out and the Intensive Care National Audit & Research Centre Case Mix Programme Database. In a sensitivity analysis only the database was used to avoid any potential double counting. - Baseline covariates were adjusted for components of the Mortality in Emergency Department Sepsis (MEDS) score

8 Economic evidence tables 8 of the protocol) - Hospital stay/icu stay - Re-admissions - Costs and QALYs were assumed to be gamma distributed, compared to normally distributed in the base case Data sources EGDT remained cost-ineffective in all sensitivity analyses. Health outcomes: Mortality data taken from the RCT (promise trial) alongside the economic evaluation. Quality-of-life weights: EQ-5D scores were elicited at 90 days, assuming an EQ-5D score of zero at randomisation, and a linear interpolation between randomisation and 90 days. Zero QALYs were assumed for people who died before 90 days. Cost sources: Costs of monitor and central venous catheter with monitoring capability was derived from the manufacturer. These costs are over 50% discount on list prices. It was assumed each site would require 2 monitors which would have a lifespan on average of 5 years. Monitor costs per patient were calculated by dividing the total costs of the monitors ( 4000) by the expected number of eligible patients over 5 years. Annual number of eligible patients calculated by taking average number of potentially eligible patients per site per year from the trial screening log data (23 patients per site per year). Some consumables sourced from hospital finance departments. Training costs per patient per hour derived from total training costs per site divided by eligible patients over 5 years. Blood products from NHS blood and transplant price list Drugs from BNF Staff costs and outpatient and community health service costs from PSSRU Hospital stay costs from NHS reference costs Comments Source of funding: NR Limitations: Adverse events not taken account of in cost effectiveness analysis (either their treatment costs or impact on QoL). Methodology behind probabilistic analysis unclear. Short time horizon. Overall applicability(d): Directly applicable Overall quality: potentially serious limitations Abbreviations: 95% CI: 95% confidence interval; CUA: cost utility analysis; da: deterministic analysis; EQ-5D: Euroqol 5 dimensions (scale: 0.0 [death] to 1.0 [full health], negative values mean worse than death); ICER: incremental cost-effectiveness ratio; NR: not reported; pa: probabilistic analysis; PSSRU: Personal Social Services Research Unit; QALYs: quality-adjusted life years (a) The paper states incremental costs and QALYs were estimated using a seemingly unrelated regression model, and they used the estimates of the means, variances and the covariance from the regression model to generate 500 estimates of incremental costs and QALYs from the joint distribution of these endpoints. By generating a cost effectiveness plane and cost effectiveness acceptability curve this implies some kind of probabilistic analysis was done but the methodology quoted isn t clear. I Monitoring 35 None.

9 Economic evidence tables 9 I Patient education, information and support 37 None. I Training and education Study Suarez Study details Population & interventions Costs Health outcomes Cost-effectiveness Economic analysis: CEA/CUA (health outcome: Life Years Gained and QALYs) Study design: Within trial analysis Approach to analysis: Preeducation program cohort (2 months before program) was compared to a post education program cohort (4 months after program). Program consisted of a 2 month educational program of training physicians and nursing staff from the emergency department, medical, and surgical wards, and ICU in early recognition of severe sepsis and the treatments in the Surviving Sepsis Campaign (SSC) protocol. Unit costs applied to prospective study data. Multivariable regression models were used to adjust for baseline Population: Patients with severe sepsis Patient characteristics: N = 2319 (b) Mean age = 62.2 (SD: 16.3) Male = 60.8% Intervention 1: Pre-intervention cohort, the 2 months prior to the educational program Intervention 2: Post intervention cohort, the 4 months following educational program. Total costs (mean per patient): Intervention 1: 14,427 Intervention 2: 15,906 Incremental (2 1): 1,479 (95% CI: NR; p=nr) Currency & cost year: 2006 Spanish Euros presented here as 2006 UK pounds (c) Cost components incorporated: Unit costs for emergency visits, surgical and medical ward daily stays, and ICU daily stays. Cost associated with the pharmacological and nonpharmacological interventions of the SSC protocol. One of the goals of the SSC protocol is QALYs (mean per patient): Intervention 1: 3.75 Intervention 2: 4.12 Incremental (2 1): 0.37 (95% CI: ; p=nr) Life Years Gained (mean per patient): Intervention 1: 5.44 Intervention 2: 5.98 Incremental (2 1): 0.54 (95% CI: ; p=nr) ICER (Intervention 2 versus Intervention 1): 5,476 per QALY gained (the adjusted ICER) (pa) 95% CI: NR Probability Intervention 2 cost-effective ( 20K threshold): 94% (read off graph) Probabilistic analysis was undertaken using non parametric bootstrapping with 2000 replications. Analysis of uncertainty: One way sensitivity analyses undertaken include: - Changing the rate for sepsis survivors from 0.51 to Making this value even more restrictive. - Quality of life weight was changed from 0.69 to The ICER was also calculated for different utility values. Only for very low utility values (lower than 0.2) was the ICER more than 20,000 (read off graph). - Discounting of Life Years Gained and

10 Economic evidence tables 10 differences of costs, QALYs, and Life Years Gained. Perspective: Spanish healthcare system perspective. Time horizon/follow-up: Post intervention cohort was a 4 month period after intervention introduced. Costs were only considered up until hospital discharge. Lifetime horizon for life years. Treatment effect duration: (a) 4 months (post intervention cohort) Discounting: Costs: NA; Health outcomes: 3% Data sources maintaining glucose control; the average cost per patient reported in a cost effectiveness analysis of insulin therapy was used. Patients who achieved the goal were applied the cost of the intensive therapy group, and patients who did not meet the goal were applied the cost of the conventional therapy group. QALYs was changed from 3% to 0%. - Discounting of Life Years Gained and QALYs was changed from 3% to 5%. - The cost of the education and training program and cost of staff time spent attending the sessions was not included in base case. These costs were included in a sensitivity analysis. All sensitivity analyses generated results similar to that of the base case. Health outcomes: Mortality and resource use data derived from a cohort before and after study (Ferrer ). Age and gender specific life expectancy for each survivor taken from the 2006 Spanish like expectancy tables. These were adjusted using the estimated reduction rate for sepsis survivors of Quality-of-life weights: The quality of lie weight used was This utility weight was obtained from a study of 6 month survivors of severe sepsis using the EQ-5D. 395 Cost sources: unit costs for emergency visits, surgical and medical ward daily stays, and ICU daily stays were from the Spanish National Health Institute. Pharmacological intervention costs from the SSC protocol were from the Spanish physician s desk reference. Non-pharmacological intervention costs were obtained from their suppliers. Insulin therapy cost was the average cost per patient from a cost effectiveness study on insulin therapy 1120 ( 144 for intensive therapy and 72 for conventional therapy). All prices in the study were adjusted to 2006 values using the Spanish consumer price index. Long term costs after discharge were not included. The costs of the training program were not included in the base case, but were included in a sensitivity analysis ( 54,270). Comments Source of funding: Supported by a grant from the Instituto de Salud Carlos III. Limitations: Only includes short term costs. Data on effectiveness from a cohort study, not RCT. Base case did not include cost if the intervention itself. Methodology not always clear; particularly around where adjusted ICER comes from. Other: The paper states that both the incremental costs and incremental QALYs/Life Years Gained were obtained by adjusting multivariable regression models to take into account possible baseline imbalances. The ICER that is reported in the study is stated to be the adjusted ICER (6,428 Euros or 5,476). It is unclear whether the adjusted ICER reported is the deterministic or probabilistic ICER, however the paper states the ICER in the text (as well as a table) then immediately in the next sentence states that

11 Economic evidence tables nearly all the bootstrap replications were below the threshold used of 30,000 euros. Thus implying this is likely to be the probabilistic ICER. Overall applicability: Partially applicable (d) Overall quality (e) Potentially serious limitations Abbreviations: CEA: cost-effectiveness analysis; 95% CI: 95% confidence interval; CUA: cost-utility analysis; EQ-5D: Euroqol 5 dimensions (scale: 0.0 [death] to 1.0 [full health], negative values mean worse than death); ICER: incremental cost-effectiveness ratio; NR: not reported; pa: probabilistic analysis; QALYs: quality-adjusted life years (a) The post intervention cohort are those that would benefit from the treatment effect of the education program. This cohort included patients during the 4 month period after the intervention. The time horizon for health outcome was lifetime so life expectancy was applied to the survivors. Therefore there is an assumption being made about the continuation of the study effect because life years will continue to vary between arms as different numbers of people will be alive in the pre and post intervention cohorts. The utility being applied to the groups is the same because the utility is the utility of sepsis survivors and is not impacted by the intervention except by the impact on mortality. (b) Note that the study this economic evaluation is based on is included in the clinical review (Ferrer2008) and the number of patients included in the study is higher than that reported here because there was also a third observation period (one year after the pre intervention group, to test the longevity of the education program) included in the clinical paper that is separate to the pre and post intervention cohorts. (c) Converted using 2006 purchasing power parities 882 (d) Directly applicable / Partially applicable / Not applicable (e) Minor limitations / Potentially serious limitations / Very serious limitations

12 GRADE tables 52 Appendix J: GRADE tables J.1 53 Scoring systems 54 None. J.2 55 Signs and symptoms 56 None. J.3 57 Blood tests 58 None. J.4 59 Lactate 60 None. J.5 61 Serum creatinine 62 None. J.6 63 Disseminated intravascular coagulation (DIC) 64 12

13 GRADE tables 65 Table 1: Disseminated intravascular coagulation (DIC) and all-cause mortality Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations DIC Control OR (95% CI) Absolute 28-day mortality - Gando observational studies very serious 1 inconsistency very serious 2 imprecision none (1.00 to 1.49) - 4 VERY LOW CRITICAL 28-day mortality - Gando observational studies very serious 1 inconsistency indirectness imprecision none (1.14 to 1.44) - 4 VERY LOW CRITICAL 28-day mortality - Ogura observational studies very serious 1 inconsistency indirectness imprecision none (1.09 to 2.75) - 4 VERY LOW CRITICAL In-hospital mortality - Gando observational studies very serious 1 inconsistency indirectness serious 3 none (1.42 to 12.59) - 4 VERY LOW CRITICAL In-hospital mortality - Gando 2007A 13

14 GRADE tables 1 observational studies very serious 1 inconsistency indirectness serious 3 none (4.54 to ) - 4 VERY LOW CRITICAL In-hospital mortality - Ogura observational studies very serious 1 inconsistency indirectness imprecision none (1.01 to 2.37) - 4 VERY LOW CRITICAL Risk of bias mainly due to the lack of evidence that physicians treating patients were blinded to the DIC status. The assumed lack of blinding means that knowledge of DIC could affect treatment, which would possibly affect outcome. 2 The majority of the evidence included an indirect population (downgraded by one increment) or a very indirect population (downgraded by two increments) 3 Downgraded by 1increment due to a very imprecise result expressed by a very wide confidence interval 4 N/A as only adjusted or unadjusted OR was provided J.7 71 Antimicrobial treatment 72 Table 2: <1 hour versus >1 hour (adult population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <1h versus >1h (multivariable analysis) Control OR (95% CI) Absolute Mortality 8 observational studies serious 1 inconsistency indirectness imprecision none - - OR 0.87 (0.81 to 0.94) - 2 VERY LOW CRITICAL Mortality - ICU setting 14

15 GRADE tables 6 observational studies serious 1 inconsistency indirectness imprecision none - - Not estimable - 2 VERY LOW CRITICAL Mortality - ED setting 2 observational studies serious 1 inconsistency indirectness serious 3 none - - Not estimable - 2 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Absolute effect not estimable as the crude event rate for the control group was not provided 3 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 3: <2 hours versus >2 hours (adult population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <2h versus >2h (multivariable analysis) Control OR (95% CI) Absolute Mortality 4 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.73 (0.51 to 1.04) - 3 VERY LOW CRITICAL Mortality - ICU setting 15

16 GRADE tables 1 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.14 (0.02 to 0.88) - 3 VERY LOW CRITICAL Mortality - ED setting 3 observational studies serious 1 inconsistency indirectness imprecision none - - OR 0.78 (0.54 to 1.12) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided 82 Table 4: <3 hours versus >3 hours (adult population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <3h versus >3h (multivariable analysis) Control OR (95% CI) Absolute Mortality 6 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.7 (0.57 to 0.86) - 3 VERY LOW CRITICAL Mortality - ICU setting 1 observational serious 1 serious 2 none - - OR 0.8 (0.6 to - 3 VERY CRITICAL 16

17 GRADE tables studies inconsistency indirectness 1.07) LOW Mortality - ED setting 5 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.62 (0.47 to 0.82) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided 87 Table 5: <4 hours versus >4 hours (adult population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <4h versus >4h (multivariable analysis) Control OR (95% CI) Absolute Mortality 2 observational studies serious 1 inconsistency indirectness very none 3/25 serious 2 (12%) 2/16 OR 1.03 (0.49 (12.5%) to 2.14) - 3 VERY LOW CRITICAL Mortality - ED setting 2 observational studies serious 1 inconsistency indirectness very none - - OR 1.03 (0.49 serious 2 to 2.14) - 3 VERY LOW CRITICAL 17

18 GRADE tables Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided Table 6: <5 hours versus >5 hours (adult population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <5h versus >5h (multivariable analysis) Control OR (95% CI) Absolute Mortality 2 observational studies serious 1 inconsistency indirectness very none - - OR 1.07 (0.24 serious 2 to 4.77) - 3 VERY LOW CRITICAL Mortality - ED setting 2 observational studies serious 1 inconsistency indirectness very none - - OR 1.07 (0.24 serious 2 to 4.77) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided Table 7: <6 hours versus >6 hours (adult population) Quality assessment No of patients Effect Quality Importance 18

19 GRADE tables No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations <6h versus >6h (multivariable analysis) Control OR (95% CI) Absolute Mortality 3 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.72 (0.58 to 0.9) - 3 VERY LOW CRITICAL Mortality - ICU setting 2 observational studies serious 1 serious 4 indirectness serious 2 none - - OR 0.79 (0.57 to 1.08) - 3 VERY LOW CRITICAL Mortality - ED setting 1 observational studies serious 1 inconsistency indirectness serious 2 none - - OR 0.67 (0.5 to 0.9) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided 4 I2=60% (p=0.11) 103 Table 8: Hourly treatment delay (ICU, adult population) Quality assessment No of patients Effect Quality Importance 19

20 GRADE tables No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Hourly treatment delay (ICU) Control OR (95% CI) Absolute In-hospital mortality 1 observational studies serious 1 inconsistency indirectness imprecision none - - OR 1.12 (1.1 to 1.14) - 2 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Absolute effect not estimable as the crude event rate for the control group was not provided 106 Table 9: Parenteral antibiotics prior to admission to hospital Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Parenteral antibiotics prior to admission to hospital (GP) Control OR (95% CI) Absolute Mortality 1 observational studies very serious 1 inconsistency indirectness very serious 2 none - - OR 0.58 (0.21 to 1.58) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided 20

21 GRADE tables 110 Table 10: <1 hour versus >1 hour (PICU, paediatric population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Parenteral antibiotics prior to admission to hospital (GP) Control OR (95% CI) Absolute Mortality 1 observational studies very serious 1 inconsistency indirectness very serious 2 none - - OR 0.6 (0.13 to 2.86) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided Table 11: <2 hours versus >2 hours (PICU, paediatric population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Parenteral antibiotics prior to admission to hospital (GP) Control OR (95% CI) Absolute Mortality 1 observational studies very serious 1 inconsistency indirectness very serious 2 none - - OR 0.41 (0.13 to 1.35) - 3 VERY LOW CRITICAL 21

22 GRADE tables Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided Table 12: <3 hours versus >3 hours (PICU, paediatric population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Parenteral antibiotics prior to admission to hospital (GP) Control OR (95% CI) Absolute Mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none - - OR 0.25 (0.08 to 0.79) - 3 VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided Table 13: <4 hours versus >4 hours (PICU, paediatric population) Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Parenteral antibiotics prior to admission to hospital (GP) Control OR (95% CI) Absolute Mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none - - OR 0.28 (0.1-3 VERY CRITICAL 22

23 GRADE tables to 0.81) LOW Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Absolute effect not estimable as the crude event rate for the control group was not provided 126 J.8 IV fluid administration Table 14: Clinical evidence profile: 6% HES versus 0.9% saline in adults with sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 6% HES versus 0.9% saline Control Relative (95% CI) Absolute 90-day mortality 1 randomised trials serious 1 inconsistency indirectness serious 2 none 248/976 (25.4%) 224/945 (23.7%) RR 1.07 (0.92 to 1.25) 17 more per 1000 (from 19 fewer to 59 more) LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 15: Clinical evidence profile: Crystalloid versus colloid plus crystalloid in adults with severe sepsis Quality assessment No of patients Effect Quality Importance 23

24 GRADE tables No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Crystalloid versus colloid + crystalloid Control Relative (95% CI) Absolute Hospital mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none 101/235 (43%) 121/258 RR 0.92 ( fewer per 1000 (from (46.9%) to 1.12) 117 fewer to 56 more) VERY LOW CRITICAL ICU mortality 1 observational studies very serious 1 inconsistency indirectness very none 72/235 serious 2 (30.6%) 99/258 RR 0.8 ( fewer per 1000 (from (38.4%) to 1.02) 146 fewer to 8 more) VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 16: Clinical evidence profile: 20% albumin versus 6% HES in adults with severe sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 20% albumin versus 6% HES Control Relative (95% CI) Absolute 28-day mortality 24

25 GRADE tables 1 randomised trials serious 1 inconsistency indirectness very none 4/30 serious 2 (13.3%) 6/26 RR 0.58 (0.18 (23.1%) to 1.83) 97 fewer per 1000 (from 189 fewer to 192 more) VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 17: Clinical evidence profile: 4% albumin versus 0.9% Sodium Chloride BP in adults with severe sepsis Quality assessment No of patients Effect No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 4% albumin versus 0.9% Sodium Chloride Control BP Relative (95% CI) Absolute Quality Importance 28-day mortality (univariate analysis) 1 randomised trials serious 1 inconsistency indirectness serious 2 none 185/603 (30.7%) 217/615 (35.3%) RR 0.87 (0.74 to 1.02) 46 fewer per 1000 (from 92 fewer to 7 more) LOW CRITICAL 28-day mortality (multivariate analysis) 1 randomised trials risk of bias inconsistency indirectness imprecision none 137/452 (30.3%) 166/467 (35.5%) OR 0.71 (0.52 to 0.97) - 3 HIGH CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 3 Adjusted odds ratio 25

26 GRADE tables Table 18: Clinical evidence profile: Albumin versus crystalloids in adults with sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Albumin versus crystalloids Control Relative (95% CI) Absolute Mortality 1 randomised trials risk of bias inconsistency serious 1 imprecision none 710/1937 (36.7%) 763/1941 RR 0.93 (0.86 (39.3%) to 1.01) 28 fewer per 1000 (from 55 fewer to 4 more) MODERATE CRITICAL Downgraded by 1 increment because of differences regarding the study population Table 19: Clinical evidence profile: Albumin versus colloids in adults with sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Albumin versus colloids Control Relative (95% CI) Absolute Mortality 1 randomised trials serious 1 inconsistency serious 2 serious 3 none 54/143 (37.8%) 58/156 (37.2%) RR 1.02 (0.76 to 1.36) 7 more per 1000 (from 89 fewer to 134 more) VERY LOW CRITICAL 26

27 GRADE tables Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment because of differences regarding the study population 3 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 20: Clinical evidence profile: Packed red blood cells (PRBC) plus EGDT versus EGDT only in adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations PRBC + EGDT versus EGDT Control Relative (95% CI) Absolute Hospital mortality 1 observational studies risk of bias inconsistency indirectness very none 14/34 serious 1 (41.2%) 20/59 RR 1.21 (0.71 (33.9%) to 2.08) 71 more per 1000 (from 98 fewer to 366 more) VERY LOW CRITICAL Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 21: Clinical evidence profile: Red blood cells (RBC) for low threshold ( 7g/dl) versus high threshold ( 9g/dl) in adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations RBC at low versus high threshold Control Relative (95% CI) Absolute 90-day mortality 1 randomised trials risk of bias inconsistency serious 1 imprecision none 216/502 (43%) 223/496 RR 0.97 (0.84 (45%) to 1.11) 13 fewer per 1000 (from 72 fewer to 49 MODERATE CRITICAL 27

28 GRADE tables more) 90-day mortality - >70 years of age 1 randomised trials risk of bias inconsistency serious 1 imprecision none 93/173 (53.8%) 98/185 (53%) RR 1.01 (0.84 to 1.23) 5 more per 1000 (from 85 fewer to 122 more) MODERATE CRITICAL 90-day mortality - 70 years or younger 1 randomised trials risk of bias inconsistency serious 1 imprecision none 123/329 (37.4%) 125/311 RR 0.93 (0.77 (40.2%) to 1.13) 28 fewer per 1000 (from 92 fewer to 52 more) MODERATE CRITICAL Intervention does not fall within the 6-hour time frame Table 22: Clinical evidence profile: 0-2 litres versus 2-4 litres of fluids in adults with severe sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 0-2L versus Control 2-4L Relative (95% CI) Absolute Hospital mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none 97/210 (46.2%) 82/186 RR 1.05 (0.84 (44.1%) to 1.3) 22 more per 1000 (from 71 fewer to 132 more) VERY LOW CRITICAL ICU mortality 28

29 GRADE tables 1 observational studies very serious 1 inconsistency indirectness serious 2 none 66/210 (31.4%) 66/186 RR 0.89 ( fewer per 1000 (from 117 (35.5%) to 1.17) fewer to 60 more) VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 23: Clinical evidence profile: 0-2 litres versus >4 litres of fluids in adults with severe sepsis Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 0-2L versus Control >4L Relative (95% CI) Absolute Hospital mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none 97/210 (46.2%) 45/100 (45%) RR 1.03 (0.79 to 1.33) 13 more per 1000 (from 94 fewer to 149 more) VERY LOW CRITICAL ICU mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none 66/210 (31.4%) 41/100 (41%) RR 0.77 (0.56 to 1.04) 94 fewer per 1000 (from 180 fewer to 16 more) VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 24: Clinical evidence profile: 2-4 litres versus >4 litres of fluids in adults with severe sepsis Quality assessment No of patients Effect Quality Importance 29

30 GRADE tables No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations 2-4L versus Control >4L Relative (95% CI) Absolute Hospital mortality 1 observational studies very serious 1 inconsistency indirectness very none 82/186 serious 2 (44.1%) 45/100 (45%) RR 0.98 (0.75 to 1.28) 9 fewer per 1000 (from 112 fewer to 126 more) VERY LOW CRITICAL ICU mortality 1 observational studies very serious 1 inconsistency indirectness serious 2 none 66/186 (35.5%) 45/100 (45%) RR 0.79 (0.59 to 1.05) 94 fewer per 1000 (from 185 fewer to 22 more) VERY LOW CRITICAL Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 25: Clinical evidence profile: High volume (20-40ml Ringer lactate/kg) versus low volume (20ml Ringer lactate/kg) in children with septic shock Quality assessment No of patients Effect No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations High volume versus low volume Control Relative (95% CI) Absolute Quality Importance Cumulative 72-hour survival 1 randomised serious 1 none 52/74 55/73 RR 0.93 ( fewer per 1000 (from MODERATE CRITICAL 30

31 GRADE tables trials inconsistency indirectness imprecision (70.3%) (75.3%) to 1.14) 173 fewer to 105 more) Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 166 J.9 Escalation of care 167 None. J Inotropic agents and vasopressors 169 Table 26: Clinical evidence profile: Norepinephrine versus vasopressin for adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine versus vasopressin Control Relative (95% CI) Absolute 28-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 150/382 (39.3%) 140/396 (35.4%) RR 1.11 (0.93 to 1.33) 39 more per 1000 MODERATE (from 25 fewer to 117 more) CRITICAL 90-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 188/379 (49.6%) 172/392 (43.9%) RR 1.13 (0.97 to 1.31) 57 more per 1000 MODERATE (from 13 fewer to 136 more) CRITICAL 31

32 GRADE tables ICU mortality 2 randomised trials serious 2 inconsistency indirectness very none 13/25 serious 1 (52%) 11/28 (39.3%) RR 1.26 (0.72 to 2.21) 102 more per 1000 (from 110 fewer to 475 more) VERY LOW CRITICAL Requiring renal replacement therapy at 48 hours 1 randomised trials serious 2 inconsistency indirectness very none 8/15 serious 1 (53.3%) 5/15 RR 1.6 (0.68 (33.3%) to 3.77) 200 more per 1000 (from 107 fewer to 923 more) VERY LOW NOT IMPORTANT New onset of tachyarrhythmias 1 randomised trials serious 2 inconsistency indirectness very none 4/15 serious 1 (26.7%) 1/15 (6.7%) RR 4 (0.5 to 31.74) 200 more per 1000 (from 33 fewer to 1000 more) VERY LOW NOT IMPORTANT Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 2 Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias Table 27: Norepinephrine versus dopamine for adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine versus dopamine Control Relative (95% CI) Absolute 32

33 GRADE tables 28-day mortality 1 randomised trials very serious 1 inconsistency indirectness serious 2 none 51/118 (43.2%) 67/134 (50%) RR 0.86 (0.66 to 1.13) 70 fewer per 1000 (from 170 fewer to 65 more) VERY LOW CRITICAL Mortality 3 randomised trials very serious 1 inconsistency indirectness serious 2 none 23/40 (57.5%) 28/40 (70%) RR 0.82 (0.59 to 1.15) 126 fewer per 1000 (from 287 fewer to 105 more) VERY LOW CRITICAL Hospital mortality 1 randomised trials very serious 1 inconsistency indirectness very serious 2 none 7/16 (43.8%) 10/16 RR 0.7 (0.36 (62.5%) to 1.37) 188 fewer per 1000 (from 400 fewer to 231 more) VERY LOW CRITICAL Incidence of arrhythmias 1 randomised trials very serious 1 inconsistency indirectness imprecision none 14/118 (11.9%) 51/134 RR fewer per 1000 (38.1%) (0.18 to 0.53) (from 179 fewer to 312 fewer) LOW NOT IMPORTANT Length of stay in the hospital (Better indicated by lower values) 1 randomised trials very serious 1 inconsistency indirectness imprecision none MD 0.7 lower (4.36 lower to 2.96 higher) LOW IMPORTANT 33

34 GRADE tables Length of stay on the ICU (Better indicated by lower values) 1 randomised trials very serious 1 inconsistency indirectness imprecision none MD 0.7 higher (1.15 lower to 2.55 higher) LOW IMPORTANT Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias 2 Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 28: Norepinephrine versus epinephrine for adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine versus epinephrine Control Relative (95% CI) Absolute 28-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 24/82 (29.3%) 17/76 RR 1.31 (22.4%) (0.76 to 2.24) 69 more per 1000 MODERATE CRITICAL (from 54 fewer to 277 more) 90-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 30/82 (36.6%) 23/74 RR 1.18 (31.1%) (0.76 to 1.83) 56 more per 1000 MODERATE CRITICAL (from 75 fewer to 258 more) Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 34

35 GRADE tables 179 Table 29: Dopexamine versus dopamine for adults with septic shock Quality assessment No of patients Effect Quality Importance No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Dopexamine versus Control dopamine Relative (95% CI) Absolute 28-day mortality 1 randomised trials risk of bias inconsistency indirectness very none 5/20 serious 1 (25%) 4/21 (19%) RR 1.31 (0.41 to 4.2) 59 more per 1000 (from 112 fewer to 610 more) LOW CRITICAL Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 30: Norepinephrine plus dobutamine versus epinephrine for adults with septic shock Quality assessment No of patients Effect No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine + dobutamine versus epinephrine Control Relative (95% CI) Absolute Quality Importance 28-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 58/169 (34.3%) 64/161 (39.8%) RR 0.86 (0.65 to 1.14) 56 fewer per 1000 MODERATE (from 139 fewer to 56 more) CRITICAL 35

36 GRADE tables 90-day mortality 1 randomised trials risk of bias inconsistency indirectness imprecision none 85/169 (50.3%) 84/161 (52.2%) RR 0.96 (0.78 to 1.19) 21 fewer per 1000 (from 115 fewer to 99 more) HIGH CRITICAL 7-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 34/169 (20.1%) 40/161 (24.8%) RR 0.81 (0.54 to 1.21) 47 fewer per 1000 MODERATE (from 114 fewer to 52 more) CRITICAL 14-day mortality 1 randomised trials risk of bias inconsistency indirectness serious 1 none 44/169 (26%) 56/161 (34.8%) RR 0.75 (0.54 to 1.04) 87 fewer per 1000 MODERATE (from 160 fewer to 14 more) CRITICAL Mortality 2 randomised trials very serious 2 inconsistency indirectness very serious 1 none 13/26 (50%) 13/26 (50%) RR 1 (0.58 to 1.71) 0 fewer per 1000 (from 210 fewer to 355 more) VERY LOW CRITICAL Mortality at discharge from ICU 1 randomised trials risk of bias inconsistency indirectness imprecision none 75/169 (44.4%) 75/161 (46.6%) RR 0.95 (0.75 to 1.21) 23 fewer per 1000 (from 116 fewer to 98 more) HIGH CRITICAL 36

37 GRADE tables Mortality at discharge from hospital 1 randomised trials risk of bias inconsistency indirectness imprecision none 82/169 (48.5%) 84/161 (52.2%) RR 0.93 (0.75 to 1.15) 37 fewer per 1000 (from 130 fewer to 78 more) HIGH CRITICAL Number of serious adverse events during catecholamine infusion 1 randomised trials risk of bias inconsistency indirectness very serious 1 none 41/169 (24.3%) 43/161 (26.7%) RR 0.91 (0.63 to 1.31) 24 fewer per 1000 (from 99 fewer to 83 more) LOW NOT IMPORTANT Number of serious adverse events after catecholamine infusion 1 randomised trials risk of bias inconsistency indirectness very serious 1 none 13/169 (7.7%) 12/161 (7.5%) RR 1.03 (0.49 to 2.19) 2 more per 1000 (from 38 fewer to 89 more) LOW NOT IMPORTANT Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs 2 Downgraded by 1 increment if the majority of the evidence was at high risk of bias, and downgraded by 2 increments if the majority of the evidence was at very high risk of bias Table 31: Norepinephrine plus dopexamine versus norepinephrine plus epinephrine for adults with septic shock Quality assessment No of patients Effect No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine + dopexamine versus epinephrine Control Relative (95% CI) Absolute Quality Importance 37

38 GRADE tables 28-day mortality 1 randomised trials risk of bias inconsistency indirectness very none 2/12 serious 1 (16.7%) 3/10 (30%) RR 0.56 (0.11 to 2.7) 132 fewer per 1000 (from 267 fewer to 510 more) LOW CRITICAL 90-day mortality 1 randomised trials risk of bias inconsistency indirectness very none 3/12 serious 1 (25%) 4/10 (40%) RR 0.62 (0.18 to 2.16) 152 fewer per 1000 (from 328 fewer to 464 more) LOW CRITICAL Downgraded by 1 increment if the confidence interval crossed one MID or by 2 increments if the confidence interval crossed both MIDs Table 32: Norepinephrine plus epinephrine versus norepinephrine plus dobutamine for adults with septic shock Quality assessment No of patients Effect No of studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations Norepinephrine + epinephrine versus norepinephrine + dobutamine Control Relative (95% CI) Absolute Quality Importance 28-day mortality 1 randomised trials risk of bias inconsistency indirectness very none 15/30 serious 1 (50%) 16/30 (53.3%) RR 0.94 (0.57 to 1.53) 32 fewer per 1000 (from 229 fewer to 283 more) LOW CRITICAL 38